5.5 KiB
Arrange reset native outcome independent review
Accept the bounded expanded/folded Arrange-reset workflow demonstrated here. No material blocker found in the reviewed outcome. This is Windows native wiring, preservation, history, source-save and reopen evidence at d6f45a13a7b8ab8dc9b84ee432a832609f5af6ed, not full issue32 layout-quality acceptance or J/L router adoption.
Reviewed all18 observation JSON notes and independently viewed all18 corresponding full-window PNGs with view_image. Read all eight checkpoint receipts/oracle outputs and both final close receipts. No GUI action, compiler, model helper or checkpoint rerun by reviewer.
Model and lifecycle evidence
Each case has after/Undo/Redo/reopened exports: eight checkpoints accepted, failure null, immutable inputs unchanged, oracle exit0 and complete_project_equal=true. Independently read the captured exported JSON and expected baseline files: all eight pairs are byte-identical, stronger than the reported typed-model equality for these particular artifacts. Verified snapshot/expected/supplemental-script/comparison-receipt hashes. Expected after-Projects remain candidate-derived UI-wiring oracles; independent core tests establish the reset contract separately.
Additional independent model inspection confirms expanded Verified report resets its two pins (610,210),(610,420) and full label override; folded Verified report retains both pins and label fraction0.5/offset(0,-32). Both cases retain route entry order, attachments, semantic port anchor records, semantic elements/relations and the entire Review sentinel diagram. Filtered edge's pin (1020,270) remains exact in this fixture; this is not a promise that all filtered routes remain at fixed world coordinates. Original working JSON imports remain byte-identical to immutable originals.
Run01 PID7812 and run02 PID4328 both have final close receipts: exit0, failure null, observed exit, immutable inputs unchanged and empty stdout/stderr. Parent observations establish normal AltF4 causes; receipts establish actual exits. Run02 close is now complete, not merely window disappearance. Both saved source projects were reopened in fresh process run02 after run01 closure, then exported and compared against the same frozen after-models.
Native observations
Expanded before/Undo images show two authored circular handles and manual label; arranged/Redo images show the selected route without those handles while automatic elbows remain. Folded images preserve collapsed presentation and visible promoted exchanges. Source-save images report all source files saved; reopen images report source project opened and retain expanded/folded state. Both sentinel-view images show retained route/label geometry, including the selected pin in expanded sentinel. Whole-model comparison verifies hidden fields that screenshots alone cannot.
The arranged view required Fit and displays71% zoom; Undo retained viewport zoom, so folded Archive partly leaves the viewport until refit. These are recorded viewing limitations, not model-history failures. Guide card13 is readable and correctly distinguishes explicit Layout's visible actual-endpoint pin/label reset from full Reset routing, retained port/attachment choices, dormant hidden/folded overrides and Undo. Search filters topics but required six scrolls to the relevant card; retain this existing discoverability limitation.
Protocol deviation and adapter incidents
The original plan assumed overwriteable JSON checkpoints. Observation03 visibly reports that the JSON destination already exists and asks for a new file. The original remained intact. The actual workflow instead used fresh JSON exports for model checkpoints, then saved expanded and folded multi-file source projects in separate locations and reopened those sources. Export did not mark source saved or consume model history; source Save supplied the persistence/reopen evidence. This is a justified workflow adjustment, not a falsely reported successful overwrite.
The supplemental export checker is explicitly marked not stage-admitted in its receipts, with its own hash and narrow equality scope; original staged tools remained the existing set. Preserve that provenance rather than describing it as original staging. A rejected click targeted a16×16 icon screenshot; full-window screenshot1 corrected targeting. Short filename typing/set_value failures and successful absolute-path typing are adapter interaction observations, not demonstrated application defects. None authorizes editing expected model bytes; no such repair is evidenced.
Binding and limits
Native manifest SHA256 eb208a3a1fa0843be4087e1797d109dc0e3dcddb3255f3ad9f536f3d98292b65.
Recorded GUI SHA256 133375d57bdecf3f3effe5ecc13ec121e1257b90838f2cdb64fb91aa9a530d1b; oracle 0792aff4cf8460c58fa2bf2f5ddbfa478523b7c9fe9b45e19ec93090191c37c1.
Supplemental checker 5bcd8a568d5318dd62217c1e0b35538b22e34403c9727eee8814a5d0ec5a899f.
Run01 close SHA256 ffe8a3cab307ecc4ad45caf027787efdb6e41449fb304fc2cadd242a7fb5daf8; run02 6abce58644e8bf2c85e133aa479f44d66eeeb0a6d59361afb9db1d996b4ed745.
No claim of general crossing/label-aware arrangement quality, arbitrary State-boundary native behavior, Process/Scenario changes, performance, Linux, full workspace/regression-suite completion, or remote publication. This trial saves newly generated multi-file sources; original authored multi-file comment/draft preservation is separately covered by desktop tests, not inferred from this UI sequence. Remote disk-full blockage still separates local acceptance from delivery.